II.
KnowledgeFabricImpl JSON
Structured · liveknowledge-fabric-impl:zep-fabric
Zep as Agent Memory Fabric json
Inspect the normalized record payload exactly as the atlas UI reads it.
{
"id": "knowledge-fabric-impl:zep-fabric",
"_kind": "KnowledgeFabricImpl",
"_file": "agent-stack/knowledge-fabric-impls/memory-service-fabrics.yaml",
"_cluster": "agent-stack",
"attributes": {
"displayName": "Zep as Agent Memory Fabric",
"description": "Zep as an agent-facing memory knowledge fabric. Persists, summarizes,\nand retrieves conversation history with entity extraction and temporal\nawareness. Enriches memories with structured entity graphs — tracking\npeople, organizations, and concepts mentioned across conversations.\nREST API and SDKs for Python and TypeScript. As a knowledge fabric,\nZep provides conversation-centric knowledge with temporal reasoning —\nunderstanding not just what was discussed but when and in what order.\n",
"knowledgeFileFormats": [
"conversation-history",
"entity-extracts",
"fact-triples"
],
"retrievalStrategy": "hybrid",
"knowledgePersistence": "memory-service",
"knowledgeScopes": [
"user",
"agent"
],
"autoExtractionSupport": true,
"notes": "Zep's temporal awareness sets it apart from simple vector stores — it\nunderstands that knowledge has a time dimension. Fact retrieval is\ntime-weighted, so recent knowledge is prioritized. Entity extraction\nbuilds a knowledge graph of entities mentioned in conversations,\nenabling relationship-aware retrieval. This makes Zep particularly\nsuited for long-running agent relationships where context evolution\nmatters.\n"
},
"outgoingEdges": [
{
"from": "knowledge-fabric-impl:zep-fabric",
"to": "layer:12-knowledge-fabric",
"kind": "realizes",
"attributes": {}
},
{
"from": "knowledge-fabric-impl:zep-fabric",
"to": "tool:zep",
"kind": "integrates_with",
"attributes": {}
},
{
"from": "knowledge-fabric-impl:zep-fabric",
"to": "memory-system:zep-service",
"kind": "uses_memory_system",
"attributes": {}
}
],
"incomingEdges": []
}